AI Tips, News & Insights

Claude is an AI assistant built by Anthropic, a company founded specifically around AI safety research. Like ChatGPT, it is a large language model capable of writing, analysis, coding, research, and conversation. But the two tools have meaningfully different strengths.

Where Claude Tends to Shine

Claude handles very long documents well — you can paste an entire report, contract, or research paper and ask detailed questions about it. It tends to be more careful about admitting uncertainty rather than confidently generating wrong answers. Many users find its writing style more natural and its reasoning more nuanced on complex topics.

Where ChatGPT Has the Edge

ChatGPT has broader third-party integrations, a larger plugin ecosystem, and stronger coding performance on many benchmarks. Its image generation through DALL-E is built in. For pure code generation, many developers still prefer it.

Which Should You Use?

For long document analysis, nuanced writing, and careful reasoning — try Claude. For coding assistance, integrations, and image generation — lean toward ChatGPT. For most general tasks, either will serve you well. Having access to both costs very little and gives you genuine flexibility.

Think of them as nested circles. AI is the largest — it describes any system designed to simulate aspects of human intelligence. Machine learning sits inside AI — it is a specific method of achieving AI through data-driven learning rather than explicit programming. Deep learning sits inside machine learning — it is a specific technique within ML that uses neural networks with many layers.

Artificial Intelligence

The broadest term. Any computer system performing tasks that would normally require human intelligence — recognising speech, translating language, making decisions, generating content. Chess computers from the 1990s were AI. So is the spam filter in your email.

Machine Learning

A method where systems learn patterns from data rather than following hand-coded rules. Instead of programming every possible outcome, you feed the system examples and it learns to generalise. The recommendation algorithm on Netflix is machine learning.

Deep Learning

A subset of machine learning using neural networks with many layers — hence deep. It excels at finding patterns in complex unstructured data like images, audio, and text. Most modern AI breakthroughs, including ChatGPT and image recognition systems, are built on deep learning.

Most people type into AI the way they type into a search engine — short, vague, minimal context. The results they get back are correspondingly shallow. The tool is capable of far more, but it needs more to work with.

Mistake 1: Too Vague

"Write something about productivity" gives the AI almost no information. Write what format? For whom? What angle? How long? What tone? Every missing detail is filled with a generic default.

Mistake 2: No Context

AI doesn't know who you are, what you're working on, or what you've already tried. Giving background changes everything. "I'm a project manager preparing for a difficult conversation with a client who missed a deadline" produces far better results than "help me write a difficult email."

Mistake 3: Accepting the First Response

The first response is a starting point, not a final product. The most effective AI users iterate — "make it shorter," "add a more concrete example," "rewrite the opening paragraph to be more direct." Each follow-up sharpens the output significantly.

Mistake 4: No Format Instruction

If you want bullet points, say so. If you want a table, ask for one. If you need exactly five items, specify five. AI will default to whatever format feels natural — which is rarely exactly what you needed.

Most meeting preparation looks the same: a quick scan of the agenda, maybe a glance at previous notes. AI makes it possible to go into any meeting significantly better prepared in a fraction of the time.

Brief Yourself on the Topic

If the meeting covers a subject you're not fully up to speed on, ask AI to explain it at the level you need. "Explain the key considerations in enterprise software procurement in plain language" takes two minutes and can save you from feeling lost in a room full of specialists.

Anticipate Questions and Objections

Prepare Sharp Questions

Good questions are often more valuable than good answers in meetings. Ask AI to generate the ten most important questions to ask in any given context. You won't use all of them — but having them ready changes how you show up.

Summarise Background Reading

If there are documents to review before the meeting, paste them into AI and ask for a summary focused on the decisions that will need to be made. Skip the narrative, get to what matters.

When you type something into ChatGPT, Claude, Gemini, or any other AI tool, that text goes somewhere. Where it goes, who can see it, and how it might be used varies significantly between tools and between consumer and enterprise versions.

Consumer vs Enterprise Versions

Most consumer AI tools — the free or standard paid tiers — may use your conversations to improve their models unless you explicitly opt out. This means what you type could, in some form, influence future model training. Enterprise versions typically offer stronger contractual guarantees: your data stays within your organisation's boundary and is not used for training.

What This Means in Practice

For general questions, brainstorming, and public information — consumer tools are fine. For anything involving client data, proprietary business information, personal health details, or confidential strategy — use your organisation's approved enterprise tool or anonymise the information before entering it.

How to Check

Most major AI tools have a privacy or data controls section in their settings. ChatGPT allows you to turn off training data collection. Claude's privacy policy is publicly available. Your organisation's IT team should be able to tell you which tools have approved data handling agreements in place.

The average professional writes dozens of emails a day. Many of them are variations of the same types — updates, follow-ups, requests, responses to difficult situations, meeting confirmations. AI handles all of these well when given clear context.

The Right Way to Use AI for Email

Don't ask AI to write emails from nothing. Give it context: who is the recipient, what is the relationship, what is the purpose, what tone is appropriate, what outcome do you want? The more context, the less editing you'll need to do.

Difficult Emails

These are where AI adds the most value — not because it writes better than you, but because starting a difficult email is the hardest part. Giving feedback, declining a request, addressing a conflict. Let AI draft a starting point and then rewrite it in your own voice. The blank page problem disappears.

The Edit Step Is Non-Negotiable

AI email drafts need a human pass before sending. Check the tone, verify any facts mentioned, make sure it sounds like you and not like a formal template. AI gives you speed. You provide the judgment and the relationship.

Understanding what AI can't do is just as important as knowing what it can. Here are ten genuine limitations that catch users off guard.

1. Know what is true. AI generates plausible text. It has no access to ground truth and cannot verify its own statements.
2. Browse the web in real time (without a specific tool enabled). Standard AI models have a knowledge cutoff and don't know what happened last week.
3. Remember previous conversations. Each session typically starts fresh. It doesn't know what you discussed yesterday.
4. Feel anything. It has no emotions, no opinions, no preferences. What looks like personality is sophisticated pattern matching.
5. Know your specific context unless you tell it. It knows nothing about your company, your team, or your situation without you providing it.
6. Be consistently reliable on numbers. AI is notoriously unreliable at arithmetic and statistical reasoning. Always verify calculations independently.
7. Predict the future. Trend extrapolation yes. Genuine foresight no.
8. Replace expert judgment. In medicine, law, finance — AI can inform but not replace the qualified human who is accountable for the outcome.
9. Understand your intent without clear communication. Vague prompts produce vague outputs. It cannot read between the lines.
10. Be 100% consistent. Ask the same question twice and you may get meaningfully different answers. This is by design — it is a feature and a limitation simultaneously.

Both Microsoft Copilot and Google Gemini are AI assistants embedded directly into productivity suites used by hundreds of millions of people. If you use either platform at work, AI is likely already available to you — you may just not be using it yet.

Microsoft Copilot

Copilot is deeply integrated into Word, Excel, PowerPoint, Outlook, and Teams. Its strongest use cases are drafting and editing documents in Word, generating presentations from notes in PowerPoint, writing and summarising email in Outlook, and capturing meeting summaries in Teams. If your work life revolves around Microsoft 365 and you are a heavy Outlook and Teams user, Copilot is almost certainly the right choice to prioritise.

Google Gemini

Gemini is integrated into Gmail, Docs, Sheets, Slides, and Meet. Its email summarisation and drafting in Gmail is excellent, and its formula generation in Sheets is a significant time-saver for non-technical users. If your team runs primarily on Google Workspace, Gemini is where your AI investment will pay off fastest.

If You Have Both

Some organisations use a mix. In that case, use Copilot for document-heavy work and meeting follow-up in Teams, and Gemini for Gmail-heavy communication and Sheets analysis. Let the task determine the tool.

The norms around AI disclosure are still forming. Different industries, organisations, and contexts have different expectations — and those expectations are changing quickly. What is considered acceptable today may be standard practice in two years or a compliance requirement in five.

When Disclosure Is Clearly Required

• Academic work — most institutions now have explicit AI use policies
• Journalism and news content — readers have a right to know
• Legal documents and professional advice — where accountability is personal and regulated
• Content explicitly represented as original human creative work

When Disclosure Is a Matter of Professional Judgment

Using AI to draft an internal report that you then substantially edited and verified? Probably no disclosure needed. Using AI to write an entire thought leadership article published under your name with minimal editing? Most would argue that requires transparency.

The Underlying Principle

AI is a tool. Using a spell-checker does not require disclosure. Using AI to research, draft, and refine an entire document that is then presented as your own independent expertise sits in different territory. The key question is not what AI did, but what the recipient reasonably expects from you and whether those expectations are being met honestly.

Whether it is a 50-page report, a long contract, a research paper, or a lengthy email thread — AI can process it and surface what matters in seconds. The process is simple but the framing makes a significant difference to the quality of the output.

Step 1: Paste the Full Text

Copy and paste the document text directly into your AI tool. Most modern AI systems can handle substantial amounts of text. If the document is very long, prioritise the most important sections.

Step 2: Give a Focused Prompt

Don't just say "summarise this." Tell the AI what you need from the summary and what you will use it for.

Step 3: Ask Follow-Up Questions

After the summary, treat the AI like a research assistant who has just read the document. "What were the specific numbers mentioned around cost?" "What objections did the author acknowledge?" "Is there anything the author seemed uncertain about?" The document is now searchable through conversation.

The traditional barriers to learning something new — finding the right course, the right teacher, the right time — are largely gone. AI can explain almost anything at almost any level, using the analogies and examples that work for you specifically.

Ask It to Explain at Your Level

This is the most powerful simple technique. Don't just ask "what is X?" — specify your background and what level of explanation you need.

Use It to Check Your Understanding

After learning something, explain it back to the AI and ask it to identify any gaps or misconceptions in your explanation. This active retrieval is one of the most effective learning techniques known — and AI makes it available on demand.

Generate Practice Questions

Ask AI to create quiz questions on anything you are studying. Then answer them without looking at your notes. Then use AI to evaluate your answers and explain where you went wrong.

The One Caution

AI can be confidently wrong, especially on niche or technical topics. For anything you will act on professionally, verify key facts with authoritative sources. Use AI to build understanding, not as a sole reference.

The AI writing tool landscape can feel overwhelming. The honest truth is that for most writing tasks, a capable general AI tool like ChatGPT or Claude will outperform specialised writing software — at a fraction of the cost or for free. Here is where specialisation genuinely adds value.

General AI Tools (ChatGPT, Claude, Gemini)

Best for: drafting, editing, rewriting, brainstorming, summarising, changing tone or format. These handle the majority of professional and creative writing tasks exceptionally well. Start here before paying for anything else.

Grammarly

Best for: real-time grammar, clarity, and tone checking as you write. Strong integration with browsers, email clients, and word processors. Less about generation and more about refinement. Worth it if you write a lot of professional communication.

Jasper

Best for: marketing teams producing high volumes of content — ads, social posts, product descriptions, landing pages. Built specifically for marketing workflows. Overkill for individual users.

Notion AI

Best for: users already working in Notion who want AI integrated into their notes, documents, and project management without switching tools.

Performance reviews have traditionally been time-consuming, inconsistent, and often dreaded by everyone involved. AI is beginning to change several parts of this process — not by replacing human judgment, but by reducing the friction around it.

How Managers Are Using AI

Many managers are using AI to draft initial performance review language, structure feedback they have already formed in their thinking, and ensure their written feedback is specific and constructive rather than vague and general. AI can take bullet point notes about an employee's performance and help shape them into clear, professional feedback that focuses on behaviours and outcomes.

How Employees Are Using AI

Employees are using AI to help write self-evaluations, articulate achievements in compelling language, and prepare for difficult performance conversations. There is nothing dishonest about using AI to help communicate what you have genuinely accomplished — the substance still has to be yours.

What Stays Human

The actual assessment. The difficult conversation. The genuine investment in someone's growth. The context and relationship that makes feedback land rather than just land on a page. AI can draft the words. Only a human can deliver them in a way that matters.

AI systems learn from data. That data was created by human beings, reflecting the world as it exists — including its inequities, its historical injustices, and its existing power structures. When an AI learns from that data, it learns the patterns in it. Including the problematic ones.

What Bias Looks Like in Practice

A hiring algorithm trained on historical hiring data will learn that certain universities, names, and backgrounds are associated with successful hires — because historically, those were the people who got hired. Not necessarily because they were the best candidates. A medical AI trained predominantly on data from one demographic may underperform on others. A facial recognition system trained on limited data sets may be significantly less accurate for people with darker skin tones.

Why This Matters for Everyday Users

Even if you are not building AI systems, you may be using ones that embed bias in less obvious ways. The AI that scores resumes. The system that determines credit. The algorithm that influences which content you see. These systems encode assumptions.

What Responsible AI Development Does

Responsible AI development actively tests for bias across demographic groups, includes diverse perspectives in the development process, and builds in mechanisms for humans to review and override AI decisions in high-stakes contexts. When evaluating AI tools for important use cases, these are the right questions to ask.

A prompt library is simply a saved collection of instructions for AI that you reuse and refine over time. The prompts that reliably produce great results for your specific work are more valuable than any single AI output — because they keep working every time you need them.

Step 1: Identify Your Recurring Tasks

List every task you do more than twice a week where you currently use AI or where AI could help. Email drafting, meeting summaries, document analysis, research questions, status reports. These are your first targets.

Step 2: Build a Prompt for Each

For each task, write a prompt that includes: your role or context, the specific task, the format you want, the tone required, and any constraints. Test it a few times and refine until it reliably produces something close to what you need.

Step 3: Store and Iterate

Keep your library somewhere accessible — a simple notes app, a Notion page, a Word document. Review it monthly and update prompts that are underperforming. Your most used prompts will improve significantly over three to six months of iteration.

The core functionality is similar — both Zoom AI Companion and Microsoft Teams Intelligent Recap offer meeting summaries, transcription, and action item capture. The differences matter in practice depending on how your team works.

Microsoft Teams Intelligent Recap

Deeply integrated with the Microsoft 365 ecosystem. Meeting summaries link directly to related documents and emails. Action items can feed into Planner and other Microsoft tools. If your team lives in the Microsoft environment — SharePoint, Planner, Outlook — the integration value is significant. The AI has context from across your organisation's Microsoft data.

Zoom AI Companion

Available to paid Zoom subscribers at no additional cost — which makes it accessible to a wider range of organisations. Strong standalone meeting summary and action capture. The team chat summarisation feature is particularly useful for catching up on missed Zoom Chat threads. Less deep integration outside the Zoom ecosystem.

The Common Ground

Both tools reduce manual note-taking. Both need human review before summaries are distributed. Both raise the same privacy consideration — inform participants that AI summarisation is active. These basics apply regardless of platform.

You do not need a computer science degree to work confidently with AI. You do need a working understanding of the vocabulary. Here are the terms worth knowing.

LLM (Large Language Model)

The technology behind ChatGPT, Claude, and Gemini. A system trained on enormous amounts of text that has learned to understand and generate human language. When someone says "AI" in a workplace context, they usually mean an LLM.

Hallucination

When an AI produces false information with apparent confidence. Not a bug — a known characteristic of how language models work. They generate probable text, not verified facts.

Prompt

The instruction or question you give the AI. Prompt quality is the primary variable in output quality.

Context Window

How much text an AI can process at once in a single conversation. Larger context windows mean you can feed in longer documents and maintain more complex conversations.

RAG (Retrieval-Augmented Generation)

A technique where an AI retrieves relevant information from a specific knowledge base before generating a response — reducing hallucination and improving accuracy for specialised topics.

Fine-Tuning

Additional training applied to a base AI model to specialise it for a specific domain or task — like training a general model specifically on medical records to improve clinical performance.

AI Agent

An AI system that can take actions autonomously — not just respond to a single prompt, but plan and execute a sequence of steps to complete a goal.

The absence of an AI policy does not mean no one is using AI — it means people are making individual decisions about what is safe and appropriate without shared context or standards. A clear policy reduces risk and removes uncertainty without slowing things down.

The Four Things a Basic AI Policy Should Cover

1. Approved tools. Which AI tools are sanctioned for work use and which are not. This is primarily a data security question — approved tools have reviewed data handling agreements; unapproved ones may not.

2. Data classification. What categories of information can and cannot be entered into AI tools. At minimum: no client personal data, no confidential commercial information, no health records in consumer tools.

3. Review requirements. AI outputs used for client deliverables, published content, or decision-making require human review and sign-off. AI is a starting point, not a final product.

4. Disclosure expectations. When team members are expected to disclose AI use — to clients, in published work, in formal documents.

The Right Tone

AI policies that treat employees as potential threats produce defensiveness and workarounds. Policies framed as enabling responsible innovation — here is what you can do confidently, here is what needs care — tend to produce better outcomes. The goal is clarity, not restriction.

AI image generators create original images from text descriptions. You describe what you want — in plain language — and the system produces an image. The quality of what is now possible would have seemed implausible to professional designers just three years ago.

The Main Tools

DALL-E (built into ChatGPT Plus) — good general-purpose image generation, easy to access if you already use ChatGPT.

Midjourney — produces consistently high-quality, artistic results. Currently accessed through Discord. Preferred by many designers and creative professionals.

Adobe Firefly — built into Adobe Creative Cloud. Strong for commercial use because it was trained on licensed content, reducing copyright risk.

Canva AI — integrated into Canva's design platform. Accessible for non-designers who are already using Canva for presentations and social content.

Practical Use Cases

• Concept illustrations for presentations and reports
• Social media visual content
• Placeholder images during design prototyping
• Visual brainstorming for creative projects

Current AI tools are largely reactive. You ask, they answer. You prompt, they respond. AI agents represent a different model — systems that are given a goal and autonomously figure out how to achieve it, taking actions along the way without requiring a human to approve each step.

What Agents Can Already Do

Early AI agents can browse the web, write and execute code, manage files, send emails, fill in forms, and chain these actions together to complete complex workflows. A research agent might be given a topic and autonomously search multiple sources, synthesise findings, and produce a formatted report. A scheduling agent might review your calendar, draft meeting proposals, and send them without being asked for each step.

What This Means for Work

Tasks that currently require a human to be in the loop for coordination — not judgment, but just logistics — become automatable. The value of human work shifts further toward tasks requiring genuine judgment, creative thinking, relationship management, and ethical reasoning.

The Skills That Matter More

Knowing what to ask agents to do. Understanding their limitations and failure modes. Reviewing and quality-checking agent outputs. Designing workflows where human and AI judgment are applied at the right stages. These are the skills that will differentiate effective workers in an agent-enabled environment.

If you asked most office workers how they use AI, you'd hear the same two answers: writing emails and searching for information. Both are great — but they barely scratch the surface of what AI can do for your daily work.

Here are five use cases that consistently deliver high value but remain underused in most workplaces.

1. Summarising Long Documents

Whether it's a 40-page report, a lengthy contract, or a meeting transcript, AI can condense it into a clear, structured summary in seconds. Paste the text and ask for the key points, decisions, or action items.

2. Preparing for Meetings

Before your next big meeting, describe the context to AI and ask it to generate likely discussion points, potential objections, or questions you should be ready to answer. It's like having a prep partner available at any hour.

3. Drafting Difficult Emails

Not just routine emails — the hard ones. Giving feedback, declining a request, following up after a missed deadline. AI can help you find the right tone when you're not sure how to start.

4. Turning Notes Into Structured Content

Dump your rough meeting notes or brainstorm ideas into AI and ask it to organise them into a clear structure — an agenda, an action list, a project brief. Raw thinking becomes polished output in minutes.

5. Learning Something New Quickly

Need to understand a topic before a meeting? Ask AI to explain it at the right level for you — with analogies, examples, and a summary you can actually use in conversation.

If you've used ChatGPT, Claude, or any large language model, there's a good chance you've encountered a hallucination without realising it. It's one of the most talked-about limitations of modern AI — and one of the most misunderstood.

What Is a Hallucination?

An AI hallucination is when a language model produces information that sounds confident, plausible, and well-written — but is factually wrong or completely made up. It might cite a study that doesn't exist, quote a person who never said it, or state a statistic that has no basis in reality.

The term "hallucination" is used because the AI isn't lying intentionally — it genuinely can't distinguish between something it "knows" and something it's generating based on patterns. It has no internal fact-checker.

Why Does It Happen?

Language models work by predicting the most statistically likely next word based on everything they've been trained on. When asked about something outside their knowledge — or when they're uncertain — they don't say "I don't know." They generate a plausible-sounding answer anyway.

How to Spot and Avoid Them

• Verify anything specific — statistics, dates, names, citations, and legal or medical information should always be checked against a reliable source
• Ask for sources — then actually check that they exist
• Be more skeptical for niche topics — AI is less reliable on very specific or recent subjects
• Cross-reference — if two different sources confirm the same thing, it's more likely to be accurate

Welcome. If you've found your way here, you're probably curious about artificial intelligence — what it actually is, how it works, and more importantly, how it fits into your everyday life and work.

That's exactly what AILiterate is for. This blog is a companion to our main learning guides — a place for shorter, more timely content: tips, observations, tool updates, and practical advice as the AI landscape continues to evolve.

What We'll Cover Here

• Practical tips for using AI tools more effectively
• Plain-language explanations of new AI developments
• Honest takes on which tools are worth your time
• Answers to questions readers ask us most often

Who This Is For

The same people our guides are written for — professionals, students, educators, and curious people who want to understand and use AI without needing a computer science degree. If that's you, you're in the right place.

We publish when we have something genuinely useful to say — not on a rigid schedule. Quality over quantity, always.

Thanks for being here. More soon.